TW201928740A - Keyword confirmation method and apparatus - Google Patents

Abstract

A keyword confirmation method and apparatus are provided. A keyword confirmation method includes: obtaining first audio data, the first audio data being recognized as a keyword; obtaining a pronunciation similarity probability of a similar pronunciation unit corresponding to at least one fragment of the first audio data and second audio data; determining that multiple contiguous silence fragments exist in second audio data contiguous in time with the first audio data; utilizing the silence probability, as well as a pronunciation similarity probability corresponding to fragment(s) of the first audio data and/or a pronunciation similarity probability corresponding to fragment(s) of the second audio data, evaluating whether the second audio data is silence; and confirming the first audio data as an effective keyword.

Description

Keyword confirmation method and device

The present application relates to the field of computers, and in particular, to a method and a device for confirming a keyword.

In the field of computer human-computer interaction, keywords are used to wake up electronic devices such as car terminals to enable human-computer interaction functions or perform certain functions, and have been used by a variety of electronic devices. For example, when the user speaks certain keywords such as starting the system and zooming in on the map, the electronic device detects the keywords of the user and executes the corresponding function of activating the system and zooming in on the map. This way of voice control greatly increases user convenience. However, how to determine the keywords sent by the user without causing misjudgment, such as identifying normal dialogs that are not keywords as keywords, or erroneously identifying keywords as non-keywords and refusing to start, has become a solution in this field. The problem. In order to solve this problem, a solution proposed in the prior art is to compare the voice issued by the user with the keywords in the keyword database, and if it can be matched, the voice command issued by the user is regarded as the keyword, and the electronic device The corresponding instruction is executed accordingly; if it fails to match, it is not a keyword and the electronic device does not execute the instruction. However, this solution is easy to identify content that is not a keyword in the user's conversation as a keyword, such as "The system does not need to be activated yet" or "I just want to know if the map zoom function is available", Keywords of this type of normal dialogue that include "start the system" or "enlarge the map" will be recognized as keywords by the electronic device, and the instruction will be executed incorrectly.

In view of the above problems, an embodiment of the present invention provides a method and a device for confirming a keyword to solve the problems existing in the prior art. In order to solve the above-mentioned problem, an embodiment of the present application discloses a keyword confirmation method, including: 第一 acquiring first audio data, the first audio data being identified as a keyword; determining a first continuous time series with the first audio data. There are multiple consecutive mute segments in the second audio material; (1) Confirm that the first audio material is a valid keyword. In order to solve the above problem, a second embodiment of the present application discloses a keyword confirmation method, including: acquiring first audio data, the first audio data being identified as a keyword; judging that the first audio data is continuous in time with the first audio data Cumulative silence probability of multiple segments of the second audio material; determining the cumulative keyword probability of multiple fragments of the first audio material; when the relationship between the cumulative silence probability and the cumulative keyword probability satisfies a second preset condition To confirm that the first audio data is a valid keyword. In order to solve the above problems, an embodiment of the present application discloses a keyword confirmation method for a vehicle terminal, including: 获取 acquiring first audio data through a vehicle audio acquisition device, where the first audio data is identified as a keyword; judging with the first There are multiple consecutive mute segments in a second audio material that is continuous in time in an audio material; confirm that the first audio material is a valid keyword, and the valid keyword is used to wake up the vehicle terminal to execute the keyword corresponding instruction. In order to solve the above problems, an embodiment of the present application discloses a keyword confirmation method for a vehicle terminal, including: 获取 acquiring first audio data through a vehicle audio acquisition device, where the first audio data is identified as a keyword; judging with the first Cumulative probability of silence of multiple segments of a second audio material continuous in audio data in time; judging the cumulative keyword probability of multiple segments of the first audio material; the relationship between the cumulative silence probability and the cumulative keyword probability When the second preset condition is satisfied, it is confirmed that the first audio data is a valid keyword, and the valid keyword is used to wake up the vehicle terminal to execute an instruction corresponding to the keyword. In order to solve the above problems, an embodiment of the present application discloses a keyword confirmation device, including: an audio data acquisition module for acquiring a first audio data, the first audio data is identified as a keyword; a silent segment determination module To determine that there are multiple consecutive silent segments in the second audio material that is temporally continuous with the first audio material; an effective keyword determination module, which is used to confirm that the first audio material is a valid keyword. In order to solve the above problems, an embodiment of the present application discloses a keyword confirmation device, including: ： an audio data acquisition module for acquiring a first audio data, the first audio data is identified as a keyword; a cumulative silence segment determination module A group for determining a cumulative silence probability of a plurality of fragments of a second audio material that is temporally continuous with the first audio material; a cumulative keyword probability determination module for determining a plurality of fragments of the first audio material Cumulative keyword probability; A valid keyword determination module configured to confirm that the first audio data is a valid keyword when the relationship between the cumulative silence probability and the cumulative keyword probability meets a second preset condition. An embodiment of the present application also discloses a terminal device including: (i) one or more processors; and one or more machine-readable media having instructions stored thereon, which when executed by the one or more processors, cause The terminal device executes the above method.一 An embodiment of the present application also discloses one or more machine-readable media having instructions stored thereon, which when executed by one or more processors, cause the terminal device to execute the above method. It can be known from the above that the keyword confirmation method provided in the embodiment of the present application includes at least the following advantages: 中 The keyword confirmation method provided in the embodiment of the present invention uses the general habit of the user, that is, before or after the keyword is issued, Silence occurs when there is a pause. By detecting whether there is silence before and after the keyword to detect whether the keyword is a valid keyword, the detection accuracy is improved, and recognition errors are avoided. At the same time, the silence of the audio data is used in the detection of silence. The detection of the fragments improves the accuracy of determining whether to mute or not, and further avoids misjudgement of keywords as non-keywords.

In the following, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art fall within the protection scope of this application. One of the core ideas of this application is to propose a keyword confirmation method that uses the silence before and after the keyword to determine whether the keyword is indeed a valid keyword. At the same time, in the detection of silence, continuous silence segments of audio data are used. Or the cumulative silence probability of multiple segments is used for judgment, and the accuracy of the judgment is improved. First Embodiment 第一 A first embodiment of the present invention proposes a keyword confirmation method. Fig. 1 is a schematic diagram of normal keywords and audio data before and after an embodiment of the present invention. As shown in Fig. 1, according to the general habits of users, there will be pauses before or after the keywords are issued, resulting in silence, so it is normal. Keywords before and after can be muted. In this way, it is possible to determine whether the voice uttered by the user is a keyword. In Figure 1, there are three possible situations for normal keywords, including: mute before keywords, mute after keywords, and mute before and after keywords. In the following, the audio material corresponding to the keyword is used as the first audio material, and the audio material corresponding to the mute part before and after the keyword is used as the second audio material. FIG. 2 is a flowchart showing steps of a keyword confirmation method according to a first embodiment of the present invention. As shown in FIG. 2, the keyword confirmation method according to the embodiment of the present invention may include, for example, the following steps: S101, obtaining first audio data, where the first audio data is identified as a keyword; In this step, an execution subject, For example, an electronic device such as a car terminal, a mobile phone, or a tablet computer can obtain audio data, which includes at least a first audio data and a front / back second audio data. The second audio material and the first audio material are continuous in time. The first audio data detected at this time has been identified as a keyword, that is, the audio data detected at this time has been confirmed to match the pre-stored keywords. As shown in FIG. 1, the electronic device can acquire and detect the first audio data corresponding to the keywords shown in FIG. 1 and the second audio data before, after, or before and after the keywords. In actual use, the electronic device A sound collection device such as a microphone can continuously capture audio. For example, audio data is acquired in units of "frames", and a frame is, for example, 10ms. After detecting the first audio data as a keyword, before acquiring the first audio data / Several frames, such as 10 frames of second audio data, for subsequent analysis. In an embodiment, it is necessary to further determine whether the first audio data is a “valid keyword”, and only when it is subsequently confirmed to be a valid keyword can a corresponding instruction be executed according to the valid keyword. S102. Determine that there are multiple consecutive mute segments in the second audio material that is temporally continuous with the first audio material. ； In this step, after inputting the segment of the second audio material into the sound unit matching model of the electronic device, The similarity with the mute unit in the sound unit library can be obtained as the mute probability of the segment. For example, for a segment of the second audio material, after matching its input sound unit model, and calculating that the similarity with the mute unit is 90%, then 90% is used as the mute probability corresponding to the segment. When this mute probability satisfies When certain requirements are met, the segment of the second audio material is considered to be a mute segment. In one embodiment, a plurality of segments may be input into a sound unit matching model of the electronic device, and the mute probability corresponding to the segment is obtained, and the mute probability is used to determine whether the segment is a mute segment.判断 After determining that the segment is a mute segment, it may be determined whether the second audio material includes multiple consecutive mute segments. For example, for multiple segments of the second audio material, after knowing whether each segment is a mute segment, it can be detected whether the mute segments are continuous mute segments. For example, each segment has an identifier f of whether it is a mute segment, When it is detected that three consecutive segments in time have the mute flag f, it is considered that there are multiple consecutive mute segments in the second audio material. (S103) Confirm that the first audio data is a valid keyword. In this step, when there are multiple consecutive mute segments in the second audio material, the second audio material is determined to be mute, thereby confirming that the keyword is a valid keyword, and subsequent execution can be performed based on the valid keyword. Corresponding instruction. For example, when the second audio material includes multiple (for example, three or more) consecutive mute segments, the second audio material is considered to be mute, and then the first audio material is determined to be a valid keyword. It is worth noting that the keywords mentioned above and below can include multiple contents. For example, the wake-up word used to wake up the operating system in the electronic device, the user's voice command, and key parameters in the command. For example, during the voice operation performed by the user on the electronic device, the input “turn on the system”, “frequency modulation to 87.6”, “87.6”, etc., all belong to the category of “keywords” proposed in the embodiments of the present invention. Not particularly limited. It can be known from the above that the keyword confirmation method provided by the first embodiment of the present invention has at least the following technical effects: 中 The keyword confirmation method provided by the embodiment of the present invention uses the general habit of the user, that is, before the keyword is issued or After that, there will be a pause to generate silence. By detecting whether there is silence before and after the keyword to detect whether the keyword is a valid keyword, the detection accuracy is improved, and recognition errors are avoided. At the same time, audio data is used in the detection of silence. The detection of the mute segment improves the accuracy of judging whether it is mute or not, and further avoids misjudgement of keywords as non-keywords. Second Embodiment 第二 A second embodiment of the present invention proposes a keyword confirmation method. FIG. 3 is a flowchart of steps in a keyword confirmation method according to a second embodiment of the present invention. As shown in FIG. 3, a keyword confirmation method according to an embodiment of the present invention includes the following steps: S201, obtaining first audio data, where the first audio data is identified as a keyword; S202, judging time with the first audio data There are multiple consecutive mute segments in the continuous second audio material; ； S203, confirm that the first audio material is a valid keyword. The above steps S201 to S203 are the same as or similar to steps S101 to S103 of the previous embodiment, and are not repeated here. This embodiment focuses on the differences from the previous embodiment. In an embodiment of the keyword confirmation method of the present invention, the step S202, that is, the step of determining that there are multiple consecutive mute segments in the second audio material that is temporally continuous with the first audio material may include the following steps: Step: S2021, determine the pronunciation similarity probability of the segment, the pronunciation similarity probability is the maximum similarity probability between the segment and multiple pronunciation units; In this step, the segment of the second audio material may be the above, for example The audio frame may also be a fragment of other units, which is not limited herein, as long as it is a fragment obtained by dividing audio data according to a specific rule, such as time, storage method, etc., all belong to the protection scope of the present invention. For example, the segment can be an audio frame of 10ms or 20ms, or an audio paragraph of 1s. The pronunciation unit can be a unit of phonemes, syllables, words, words, etc. obtained according to the user's pronunciation. For example, when the user issues a "zebra", the pronunciation unit is divided by phonemes, and the phonemes are smaller than the syllables. From the existing phoneme set, we can know that the phoneme "b a_h nn_h m a_l a_l" corresponding to "Zebra"; when the pronunciation unit is divided by syllables, the syllable corresponding to "Zebra" is "b an ma"; When the pronunciation unit is divided by the pronunciation of the word, the corresponding division of "Zebra" is "ban ma"; when the pronunciation unit is divided by the word, the corresponding division of "Zebra" is "banma". For each division, a corresponding sound unit library can be constructed. In addition to the pronunciation unit mentioned above, the thesaurus may also include a mute unit and the like. When a segment of the second audio material has the highest probability of similarity with one of the pronunciation units pre-stored in the sound unit library, it is considered that the segment matches the pronunciation unit, then the pronunciation unit is regarded as a similar pronunciation unit, and the segment and the The similarity rate of similar pronunciation units is used as the pronunciation similarity probability. For the fragment in the second audio data, the fragment is input to the sound unit matching model of the electronic device for judgment. If the fragment has the highest similarity with the fifth pronunciation unit in the sound unit library, it is 80%, and it is the same as the sixth pronunciation unit. The second similarity is 70%, and the fifth pronunciation unit with 80% similarity can be recorded as the similar pronunciation unit corresponding to the segment, and the similarity probability of pronunciation is 80% for subsequent processing. S2022: Determine the mute probability of the segment, where the mute probability is a similar probability between the segment and the mute unit; In this step, after the segment of the second audio data is input into the sound unit matching model of the electronic device, it can be learned Its similarity to the mute unit in the sound unit library is used as the mute probability of the segment. For example, for a segment of the second audio material, after the input sound unit is matched with the model, the similarity between the second audio material and the mute unit is calculated to be 90%, and 90% is used as the mute probability corresponding to the segment. It is worth noting that the above-mentioned mute unit can be pre-stored in the sound unit library, which can be obtained through repeated training of a large amount of data, such as comprehensive consideration of sound energy, environmental noise (including wind sound, music sound, car whistle, etc.) Waiting for the mute unit is not limited to absolute silence. The length, attributes, etc. of the mute unit may correspond to the invention unit. For example, when the pronunciation unit is divided according to phonemes, the mute unit may be a mute phoneme; when the pronunciation unit is divided according to syllables, the mute unit may be a mute syllable, which is not limited herein. S2023: When the relationship between the pronunciation similarity probability and the silence probability satisfies a preset condition, determine that the segment is a silence segment. The preset conditions include, for example: The absolute value of the difference between the pronunciation similarity probability and the silence probability of the segment is less than the first threshold. In this step, the pronunciation similarity probability and the corresponding mute probability corresponding to the previously obtained segment of the second audio material may be used to determine whether the segment of the second audio material is muted. As can be seen from the above, in the solution proposed by the embodiment of the present invention, the judgment of the silence probability is not to compare the segments of the audio data with absolute silence, but to compare the similarity probability of pronunciation with the corresponding silence probability, comprehensively considering the environment noise Therefore, the solution provided by the present invention can avoid rejecting correct keywords due to inaccurate silence judgment. There are many ways to judge whether a segment of an audio material is muted by using the probability of sound similarity and the probability of silence, which is described here by way of example. For example, the segment satisfies the requirement that the absolute value of the difference between the pronunciation similarity probability pmax (indexframe) and the silence probability psil (indexframe) is less than 15%, namely:The segment is considered to be a silent segment. In an embodiment of the keyword confirmation method of the present invention, in the above sub-step S2022, the judgment of the silence probability may also be the similarity probability of the pronunciation unit and the silence unit corresponding to the similarity probability of the pronunciation unit and the silence unit corresponding to the maximum similarity probability. To judge. That is, the sub-step S2022 may be replaced with the following sub-steps: S2024: determine the mute probability of the segment, where the mute probability is a similarity probability between the pronunciation unit and the mute unit corresponding to the maximum similarity probability. In substep S2021, the pronunciation similarity probability of the segment has been determined. For example, in the foregoing example, the segment in the second audio data is determined through the sound unit matching model of the electronic device, and the segment and the fifth pronunciation in the sound unit library are determined. The unit has the highest similarity, which is 80%, and the fifth pronunciation unit corresponding to the 80% maximum similarity probability here is a similar pronunciation unit. In this sub-step, the similarity probability of the fifth sounding unit and the mute unit can be calculated as the mute probability of the segment. According to the methods listed above and the technical capabilities of those skilled in the art, those skilled in the art can use these probabilistic similarity probabilities and mute probabilities to set many judgment methods to determine whether the second audio data segment is muted. The present invention is not particularly limited . After step S2023 or step S2024, that is, after determining that the segment is a mute segment, the step S202, that is, determining that there are multiple consecutive mute segments in a second audio material that is temporally continuous with the first audio material The step may further include the following sub-steps: S2025: Determine, according to the determined mute segment, that there are multiple consecutive mute segments in the second audio material. In this step, it can be determined whether the second audio material includes multiple consecutive mute segments. For example, for multiple segments of the second audio material, for example, in a sub-step S2023 or S2024, an identifier f for each segment is set as a mute segment. When it is detected that three consecutive segments in time have a mute flag f, It is considered that there are multiple consecutive mute segments in the second audio material. In an embodiment of the keyword confirmation method of the present invention, in the above sub-step S2025, that is, in the step of determining that there are multiple consecutive silent segments in the second audio material, the "multiple" may be three or more That is, sub-step S2025 may be: determine that there are more than three consecutive mute segments in the second audio material. In an embodiment of the keyword confirmation method of the present invention, before step S201, that is, the step of obtaining audio data, the method may further include: S200, detecting whether the collected audio data includes keywords. In this step, multiple keywords can be pre-stored in the keyword library of the electronic device, such as "hello zebra", "open system", "zoom in map", "zoom out map", "exit navigation" and so on. The keywords in the first audio material may be any of them. You can use the keyword thesaurus to calculate the probability that the input first audio material is similar to these keywords, and select the words with the highest probability and higher than the set threshold. As the detected keywords. Specifically, for example, the sound unit matching method of the present invention can be used to divide the audio data into multiple segments. When a segment has the highest similarity probability with one of the pronunciation units pre-stored in the sound unit library, the segment is considered to be similar to the When the sound unit matches, the sound unit is regarded as a similar sound unit, and the similarity rate between the segment and the similar sound unit is used as the sound similarity probability.针对 In the case of a piece of speech, such as the first audio material, the probabilities of pronunciation similarity of multiple segments are processed. For example, after multiplication, the path with the highest probability is obtained, and the word corresponding to the path is used as the matching keyword. In an embodiment of the keyword confirmation method of the present invention, the keyword has attribute information, and the step S203, the step of confirming that the first audio material is a valid keyword may include: when the attribute information of the keyword When the main keyword and the second audio material before the keyword are muted, confirm that the keyword is a valid main keyword. In the embodiment of the present invention, each keyword may correspond to an attribute information, and the information records whether the keyword is a primary keyword or a secondary keyword. The multiple keywords pre-stored in the keyword library of the electronic device can be categorized as, for example, primary and secondary keywords, such as "hello zebra", "open system", etc., and set as the primary keywords, "zoom in map", "zoom out" Map, "" exit navigation, "and so on. For the main keyword, if you consider that there may be no content after the main keyword, or it may directly recognize the voice, such as "Hello Zebra, please help me find the way to Zhongguancun", you can set the audio information before the keyword as If the keyword is muted and the attribute information of the keyword is the main keyword, confirm that the keyword is the main keyword; do not check whether the keyword is muted after the keyword. In an embodiment of the keyword confirmation method of the present invention, the keyword has attribute information, and the keyword has attribute information. The step S203, the step of confirming that the first audio data is a valid keyword may include: When the attribute information of the keyword is a sub-keyword, and the second audio data before and after the keyword are both mute, confirm that the keyword is a valid sub-keyword. In this step, the sub-keyword may be a command that the user asks the electronic device to execute directly, such as "zoom in the map". You can set the content before and after the keyword to be mute, and the attribute of the keyword is a sub-keyword, and then confirm that the keyword is a sub-keyword. When the user says "I just want to try to enlarge the map", "I do n’t know if I can enlarge the map" or "Enlarge the map", the keywords can be detected, but they are not satisfied The condition of silence will not be judged as a valid keyword. In summary, the keyword confirmation method provided in this embodiment has at least the following advantages: 中 The keyword confirmation method applied to a vehicle terminal according to the embodiment of the present invention utilizes the general habits of users, that is, There will be pauses before and after to generate silence. By detecting whether there is silence before and after the keyword to detect whether the keyword is a valid keyword, the detection accuracy is improved, and recognition errors are avoided. At the same time, in the detection of silence, the use of Detection of mute segments of audio materials improves the accuracy of determining whether to mute or not, and further avoids misjudgement of keywords as non-keywords. In addition, the keyword confirmation method proposed in this embodiment includes at least the following advantages: 的 In an optional embodiment of the keyword confirmation method provided in the embodiment of the present invention, a preferred judgment method is proposed, which uses the second audio data The ratio of the cumulative probability of silence to the cumulative keyword probability of the first audio material and / or the difference between the similarity probability of the pronunciation of the second audio material and the probability of silence, to determine whether the second audio material is muted, so that the judgment result is more accurate; Different types of keywords—main keywords and subsidiary keywords—provide different ways of further confirmation, making the judgment results more reliable. Third Embodiment 第三 A third embodiment of the present invention proposes a keyword confirmation method. FIG. 5 is a flowchart of steps in a keyword confirmation method according to a third embodiment of the present invention. As shown in FIG. 5, the keyword confirmation method according to the embodiment of the present invention includes the following steps: S301, acquiring first audio data, where the first audio data is identified as a keyword; In this step, the execution subject, for example, a vehicle Electronic devices such as a terminal, a mobile phone, and a tablet computer can obtain audio data, which includes at least a first audio data and a front / back second audio data. The second audio material and the first audio material are continuous in time. The first audio data detected at this time has been identified as a keyword, that is, the audio data detected at this time has been confirmed to match the pre-stored keywords. The electronic device can acquire and detect the first audio data corresponding to the keyword and the second audio data before, after, or before and after the keyword. In actual use, the sound collection device of the electronic device such as a microphone can continuously collect audio, audio The data is acquired in units of "frame", for example, a frame is 10ms, and after detecting the first audio material as a keyword, several frames before / after the first audio material are acquired, such as 10 frames of the second audio Data for subsequent analysis. S302. Determine a cumulative mute probability of a plurality of segments of the second audio material that is continuous with the first audio material in time; The cumulative mute probability p (sil) of the plurality of segments of the second audio material may use the second audio material. The product or sum of the silence probabilities of each segment is obtained. For each segment, the mute probability can be calculated in the manner provided in the foregoing first and second embodiments. In step S302, the mute probability can be obtained by adding or multiplying the mute probability. S303: Determine the cumulative keyword probability of multiple segments of the first audio material; In this step, the cumulative keyword probability p (kws) of the first audio material may be corresponding to multiple segments of the first audio material The product of the probabilities of pronunciation similarity; for example, for the first segment and the second segment of the first audio material, these two segments are input into the sound unit matching model of the electronic device for judgment. The similarity of the pronunciation unit is the highest, which is 90%, and the similarity with the second pronunciation unit is 80%; then the first pronunciation unit is used as the similar pronunciation unit of the first segment, and 90% is used as the pronunciation similarity probability of the first segment; The pronunciation similarity probability of the second segment is, for example, 70%, and the cumulative keyword probability of the first audio material is 80% × 70%. It is worth noting that the cumulative keyword probability p (kws) of the first audio material can be obtained from the pre-processing, and is not limited to the instant calculation. The invention does not limit the way of obtaining it. S304: When the relationship between the cumulative silence probability and the cumulative keyword probability satisfies a second preset condition, confirm that the first audio material is a valid keyword. In an embodiment of the keyword confirmation method of the present invention, the second preset condition includes: The ratio of the cumulative silence probability to the cumulative keyword probability is greater than a second threshold.这一 In this embodiment, for example, the larger the ratio of p (sil) / p (kws), the more accurate the judgment of the second audio data as mute. Therefore, a second preset condition may be set including that the ratio of the cumulative silence probability to the cumulative keyword probability is greater than the second threshold, and when p (sil) / p (kws) is greater than the second threshold (for example, 1. 5) When the second audio data is considered to be muted.一 In an optional embodiment of the present invention, before step S301, that is, the step of obtaining audio data, the method further includes: S300, detecting whether the collected audio data includes keywords. In this step, multiple keywords can be pre-stored in the keyword library of the electronic device, such as "hello zebra", "open system", "zoom in map", "zoom out map", "exit navigation" and so on. The keywords in the first audio material may be any of them. You can use the keyword thesaurus to calculate the probability that the input first audio material is similar to these keywords, and select the words with the highest probability and higher than the set threshold. As the detected keywords. Specifically, for example, the sound unit matching method of the present invention can be used to divide the audio data into multiple segments. When a segment has the highest similarity probability with one of the pronunciation units pre-stored in the sound unit library, the segment is considered to be similar to the When the sound unit matches, the sound unit is regarded as a similar sound unit, and the similarity rate between the segment and the similar sound unit is used as the sound similarity probability.针对 In the case of a piece of speech, such as the first audio material, the probabilities of pronunciation similarity of multiple segments are processed. For example, after multiplication, the path with the highest probability is obtained, and the word corresponding to the path is used as the matching keyword. In an optional embodiment of the present invention, the keyword has attribute information, and when the relationship between the cumulative silence probability and the cumulative keyword probability satisfies a second preset condition, confirming that the first audio data is valid The step of keywords includes: When the attribute information of the keywords is the main keyword, and the second audio data before the keywords is muted, confirm that the keywords are valid main keywords. In the embodiment of the present invention, each keyword may correspond to an attribute information, and the information records whether the keyword is a primary keyword or a secondary keyword. The multiple keywords pre-stored in the keyword library of the electronic device can be categorized as, for example, primary and secondary keywords, such as "hello zebra", "open system", etc., and set as the primary keywords, "zoom in map", "zoom out Map, "" exit navigation, "and so on. For the main keyword, if you consider that there may be no content after the main keyword, or it may directly recognize the voice, such as "Hello Zebra, please help me find the way to Zhongguancun", you can set the audio information before the keyword as If the keyword is muted and the attribute information of the keyword is the main keyword, confirm that the keyword is the main keyword; do not check whether the keyword is muted after the keyword. In an optional embodiment of the present invention, the keyword has attribute information, and when the relationship between the cumulative silence probability and the cumulative keyword probability satisfies a second preset condition, confirming that the first audio data is valid The keyword step includes: When the attribute information of the keyword is a sub-keyword, and the second audio data before and after the keyword are both mute, confirm that the keyword is a valid sub-keyword. In this step, the sub-keyword may be a command that the user asks the electronic device to execute directly, such as "zoom in the map". You can set the content before and after the keyword to be mute, and the attribute of the keyword is a sub-keyword, and then confirm that the keyword is a sub-keyword. When the user says "I just want to try to enlarge the map", "I do n’t know if I can enlarge the map" or "Enlarge the map", the keywords can be detected, but they are not satisfied The condition of silence will not be judged as a valid keyword. In summary, the keyword confirmation method provided in this embodiment has at least the following advantages: 中 The keyword confirmation method provided in the embodiment of the present invention uses the general habit of the user, that is, before or after the keyword is issued, The silence is caused by the pause. By detecting whether there is silence before and after the keyword to detect whether the keyword is a valid keyword, the detection accuracy is improved, and recognition errors are avoided. At the same time, the silence segment of the audio data is used in the detection of silence. Detection improves the accuracy of judging whether it is muted or not, and further avoids misjudgement of keywords as non-keywords. In addition, the keyword confirmation method proposed in this embodiment includes at least the following advantages: 的 In an optional embodiment of the keyword confirmation method provided in the embodiment of the present invention, a preferred judgment method is proposed, which uses the second audio data The ratio of the cumulative probability of silence to the cumulative keyword probability of the first audio material and / or the difference between the similarity probability of the pronunciation of the second audio material and the probability of silence, to determine whether the second audio material is muted, making the judgment result more accurate; Different types of keywords—main keywords and subsidiary keywords—provide different ways of further confirmation, making the judgment results more reliable. Fourth Embodiment 第四 A fourth embodiment of the present invention proposes a keyword confirmation method for a vehicle terminal. FIG. 6 is a schematic diagram of a vehicle-mounted terminal including a vehicle environment. FIG. 7 is a flowchart of a keyword confirmation method for a vehicle-mounted terminal according to a fifth embodiment of the present application. As shown in FIG. 6, the vehicle includes an in-vehicle terminal 200 provided in the vehicle. The in-vehicle terminal 200 includes a speaker 400 and a microphone 700, and may further include a screen, buttons, etc. (not shown). In addition to being integrated into the vehicle terminal, the speaker 400 may also be provided in other positions inside the vehicle for the occupant 600 to listen to information. The in-vehicle terminal 200 has a calculation processing function, which can be installed with an operating system and an application program, and can also be remotely connected to the server 300 through the Internet 500 for data interaction. As shown in FIG. 7, the method for confirming a keyword of a vehicle-mounted terminal according to an embodiment of the present invention is as follows: 401 S401, acquiring first audio data through a vehicle-mounted audio acquisition device, where the first audio data is identified as a keyword; at this step In the vehicle-mounted terminal 200, audio data can be acquired, which includes at least a first audio data and a front / rear second audio data. The second audio material and the first audio material are continuous in time. The first audio data detected at this time has been identified as a keyword, that is, the audio data detected by the microphone 700 at this time has been confirmed to match the pre-stored keywords. As shown in FIG. 1, the in-vehicle terminal 200 can acquire and detect the first audio data corresponding to the keywords shown in FIG. 1 and the second audio data before, after, or before and after the keywords. In actual use, the on-vehicle The terminal's sound collection device such as a microphone can continuously collect audio. For example, audio data is acquired in units of "frames", and a frame is, for example, 10 ms. After detecting the first audio data as a keyword, the first audio data is acquired. Several frames before / after, such as 10 frames of second audio data, are subsequently analyzed. S402. It is determined that there are multiple consecutive mute segments in the second audio material that is temporally continuous with the first audio material. In this step, after inputting the segment of the second audio material into the sound unit matching model of the vehicle terminal, The similarity with the mute unit in the sound unit library can be obtained as the mute probability of the segment. For example, for a segment of the second audio material, after matching its input sound unit model, and calculating that the similarity with the mute unit is 90%, then 90% is used as the mute probability corresponding to the segment. When this mute probability satisfies When certain requirements are met, the segment of the second audio material is considered to be a mute segment. In one embodiment, multiple segments can be input into the sound unit matching model of the vehicle terminal, and the mute probability corresponding to the segment is obtained, and the mute probability is used to determine whether the segment is a mute segment. (S403) Confirm that the first audio data is a valid keyword, and the valid keyword is used to wake up the vehicle terminal to execute an instruction corresponding to the keyword. In this step, when there are multiple consecutive mute segments in the second audio material, when it is determined that the second audio material is mute, the keyword can be confirmed to be a valid keyword, and the corresponding can be performed according to the valid keyword in the future. Instructions. For example, for the multiple segments of the second audio material, it is known whether each of the segments is a mute segment. In this step, it is possible to detect whether the mute clips are continuous mute clips. When a plurality of (for example, three or more) continuous mute clips are included, the second audio data is considered to be mute, and the first audio data is determined to A valid keyword, wherein the valid keyword is used to wake up a vehicle-mounted terminal to execute an instruction corresponding to the keyword. In summary, the keyword confirmation method for an in-vehicle terminal provided in this embodiment has at least the following advantages: 中 The keyword confirmation method applied to an in-vehicle terminal according to the embodiment of the present invention uses the general habit of the user, that is, There is a pause before or after the keyword to generate silence. By detecting whether there is silence before and after the keyword to detect whether the keyword is a valid keyword, the detection accuracy is improved, and recognition errors are avoided. Using the detection of mute segments of audio data, the accuracy of determining whether to mute is improved, and further avoiding the misjudgment of keywords as non-keywords. Fifth Embodiment 第五 A fifth embodiment of the present invention provides a keyword confirmation method for a vehicle-mounted terminal. FIG. 8 is a flowchart of a keyword confirmation method for a vehicle-mounted terminal according to a sixth embodiment of the present application. As shown in FIG. 8, the method for confirming a keyword of an in-vehicle terminal according to an embodiment of the present invention is as follows: S501: Obtain a first audio material through an in-vehicle audio acquisition device, and the first audio material is identified as a keyword; In the vehicle-mounted terminal, audio data can be obtained, which includes at least a first audio data and a front / rear second audio data. The second audio material and the first audio material are continuous in time. The first audio data detected at this time has been identified as a keyword, that is, the audio data detected at this time has been confirmed to match the pre-stored keywords. The vehicle-mounted terminal can acquire and detect the first audio data corresponding to the keyword, and the second audio material before, after, or before and after the keyword. In actual use, the sound collection device of the vehicle terminal such as a microphone can continuously collect audio, audio The data is acquired in units of "frame", for example, a frame is 10ms, and after detecting the first audio material as a keyword, several frames before / after the first audio material are acquired, such as 10 frames of the second audio Data for subsequent analysis. S502. Determine a cumulative mute probability of a plurality of segments of the second audio material that is continuous with the first audio material in time; The cumulative mute probability p (sil) of the plurality of segments of the second audio material may use the second audio material. The product or sum of the silence probabilities of each segment is obtained. For each segment, the mute probability can be calculated in the manner provided in the foregoing first and second embodiments. In step S502, the mute probability can be obtained by adding or multiplying the mute probability. S503. Determine the cumulative keyword probability of multiple segments of the first audio material. In this step, the cumulative keyword probability p (kws) of the first audio material may be corresponding to multiple segments of the first audio material. The product of the probabilities of pronunciation similarity; for example, for the first segment and the second segment of the first audio material, these two segments are input into the sound unit matching model of the vehicle terminal for judgment. The similarity of the pronunciation unit is the highest, which is 90%, and the similarity with the second pronunciation unit is 80%; then the first pronunciation unit is used as the similar pronunciation unit of the first segment, and 90% is used as the pronunciation similarity probability of the first segment; The pronunciation similarity probability of the second segment is, for example, 70%, and the cumulative keyword probability of the first audio material is 80% × 70%. S504. When the relationship between the cumulative silence probability and the cumulative keyword probability satisfies a second preset condition, confirm that the first audio data is a valid keyword, and the valid keyword is used to wake the vehicle terminal to execute the key. Word corresponding instruction.这一 In this embodiment, for example, the larger the ratio of p (sil) / p (kws), the more accurate the judgment of the second audio data as mute. Therefore, a second preset condition may be set including that the ratio of the cumulative silence probability to the cumulative keyword probability is greater than the second threshold, and when p (sil) / p (kws) is greater than the second threshold (for example, 1. 5) When the second audio data is considered to be muted. In an embodiment of the keyword confirmation method of the present invention, the second preset condition includes: The ratio of the cumulative silence probability to the cumulative keyword probability is greater than a second threshold. In summary, the keyword confirmation method for an in-vehicle terminal provided in this embodiment has at least the following advantages: 中 The keyword confirmation method applied to an in-vehicle terminal according to the embodiment of the present invention uses the general habit of the user, that is, There is a pause before or after the keyword to generate silence. By detecting whether there is silence before and after the keyword to detect whether the keyword is a valid keyword, the detection accuracy is improved, and recognition errors are avoided; meanwhile, in the detection of silence, Using the detection of mute segments of audio data, the accuracy of determining whether to mute is improved, and further avoiding the misjudgment of keywords as non-keywords. It is worth noting that although the above-mentioned fourth and fifth embodiments propose a keyword confirmation method for a vehicle terminal, those skilled in the art can clearly understand that the keyword confirmation method proposed by the present invention may not be limited to a vehicle terminal, and It can be applied to various other smart devices. Such as mobile phones, servers, smart home hardware and other smart devices with computing and processing functions. Smart home hardware includes, for example, microwave ovens, ovens, washing machines, dishwashers, air conditioners, routers, smart speakers, TVs, refrigerators, vacuum cleaners, and the like.一个 In one embodiment, when the above keyword confirmation method is applied to a smart speaker, the above keywords may include, for example, "play music" or "next song". When the smart speaker receives the keyword "play music" (first audio data) through the sound receiving device, the smart speaker judges whether the second audio data is muted, and then confirms that the keyword is a valid keyword. In an embodiment, after being confirmed as a valid keyword, the smart speaker may start executing an instruction for playing music corresponding to the valid keyword. The above embodiments are described by taking a smart speaker as an example, but those skilled in the art can clearly understand that the keyword confirmation method provided by the present invention can be applied to various smart devices, which is not limited herein. Sixth Embodiment 第六 A sixth embodiment of the present invention provides a keyword confirmation device. FIG. 9 is a block diagram of a keyword confirmation device according to a seventh embodiment of the present application. As shown in FIG. 9, a keyword confirmation device according to an embodiment of the present invention includes: an audio data acquisition module 601, configured to acquire a first audio data, the first audio data is identified as a keyword; a silent segment determination module 602 To determine that there are multiple consecutive mute segments in the second audio material that is temporally continuous with the first audio material; effective keyword determination module 603, which is used to confirm that the first audio material is a valid keyword. In summary, the keyword confirmation device provided in this embodiment has at least the following advantages: 中 In the keyword confirmation device provided in the embodiment of the present invention, the general habit of the user is used, that is, before or after the keyword is issued, The silence is caused by the pause. By detecting whether there is silence before and after the keyword to detect whether the keyword is a valid keyword, the detection accuracy is improved, and recognition errors are avoided. At the same time, the silence segment of the audio data is used in the detection of silence. Detection improves the accuracy of judging whether it is muted or not, and further avoids misjudgement of keywords as non-keywords. Seventh Embodiment 第七 A seventh embodiment of the present invention provides a keyword confirmation device. FIG. 10 is a block diagram of a keyword confirmation apparatus according to an eighth embodiment of the present application. As shown in FIG. 10, a keyword confirmation device according to an embodiment of the present invention includes: an audio data acquisition module 701, configured to acquire a first audio data, where the first audio data is identified as a keyword; a cumulative mute segment determination module 702, configured to determine a cumulative silence probability of a plurality of segments of a second audio material that is temporally continuous with the first audio material; a cumulative keyword probability determination module 703, configured to determine a plurality of the first audio materials The cumulative keyword probability of the segment; A valid keyword determination module 704 is configured to confirm that the first audio data is a valid keyword when the relationship between the cumulative silence probability and the cumulative keyword probability meets a second preset condition. In summary, the keyword confirmation device provided in this embodiment has at least the following advantages: 中 In the keyword confirmation device provided in the embodiment of the present invention, the general habit of the user is used, that is, before or after the keyword is issued, The silence is caused by the pause. By detecting whether there is silence before and after the keyword to detect whether the keyword is a valid keyword, the detection accuracy is improved, and recognition errors are avoided. At the same time, the silence segment of the audio data is used in the detection of silence Detection improves the accuracy of judging whether it is muted or not, and further avoids misjudgement of keywords as non-keywords. As for the device embodiment, since it is basically similar to the method embodiment, it is described relatively simply. For relevant parts, refer to the description of the method embodiment. FIG. 11 is a schematic diagram of a hardware structure of a terminal device according to an embodiment of the present application. As shown in FIG. 11, the terminal device may include an input device 90, a processor 91, an output device 92, a memory 93, and at least one communication bus 94. The communication bus 94 is used to implement a communication connection between the components. The memory 93 may include high-speed RAM memory, and may also include non-volatile storage NVM, such as at least one magnetic disk memory. The memory 93 may store various programs for performing various processing functions and implementing the method of this embodiment. step. Optionally, the processor 91 may be, for example, a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), or a programmable Design logic device (PLD), field programmable gate array (FPGA), controller, microcontroller, microprocessor, or other electronic components, the processor 91 is coupled to the input device 90 and output through a wired or wireless connection Equipment 92. Optionally, the input device 90 may include multiple input devices, for example, it may include at least one of a user-oriented user interface, a device-oriented device interface, a software-programmable interface, a camera, and a sensor. Optionally, the device-oriented device interface may be a wired interface for data transmission between the device and a hardware insertion interface (for example, a USB interface, a serial port) for data transmission between the device and the device. Optionally, the user-oriented user interface may be, for example, a user-oriented control button, a voice input device for receiving voice input, and a touch-sensing device (such as Touch screen, touchpad, etc.); optionally, the programmable interface of the above software can be, for example, an entry for a user to edit or modify a program, such as an input pin interface or input interface of a chip; Optionally, the above-mentioned transceiver may be a radio frequency transceiver chip having a communication function, a baseband processing chip, and a transceiver antenna. A voice input device such as a microphone can receive voice data. The output device 92 may include output devices such as a display and an audio. In this embodiment, the processor of the terminal device includes functions for executing each module of the data processing device in each device. For specific functions and technical effects, reference may be made to the foregoing embodiments, and details are not described herein again. FIG. 12 is a schematic diagram of a hardware structure of a terminal device according to another embodiment of the present application. FIG. 12 is a specific embodiment of the implementation process of FIG. 11. As shown in FIG. 12, the terminal device in this embodiment includes a processor 101 and a memory 102. The processor 101 executes the computer program code stored in the memory 102 to implement the methods in FIG. 1 to FIG. 7 in the foregoing embodiment. The memory 102 is configured to store various types of data to support operation at the terminal device. Examples of these materials include instructions for any application or method used to operate on the terminal device, such as messages, pictures, videos, etc. The memory 102 may include random access memory (RAM), and may also include non-volatile memory (non-volatile memory), such as at least one magnetic disk memory. Optionally, the processor 101 is provided in the processing component 100. The terminal device may further include a communication element 103, a power source element 104, a multimedia element 105, an audio element 106, an input / output interface 107, and / or a sensor element 108. The specific components and the like included in the terminal device are set according to actual requirements, which is not limited in this embodiment. The processing unit 100 generally controls the overall operation of the terminal device. The processing element 100 may include one or more processors 101 to execute instructions to complete all or part of the steps of the methods in FIG. 1 to FIG. 7 described above. In addition, the processing element 100 may include one or more modules to facilitate interaction between the processing element 100 and other elements. For example, the processing element 100 may include a multimedia module to facilitate the interaction between the multimedia element 105 and the processing element 100. The power source element 104 supplies power to various elements of the terminal device. The power supply element 104 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for end devices. The multimedia element 105 includes a display screen that provides an output interface between the terminal device and the user. In some embodiments, the display screen may include a liquid crystal display (LCD) and a touch panel (TP). If the display screen includes a touch panel, the display screen can be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touch, swipe, and gestures on the touch panel. The touch sensor may not only sense a boundary of a touch or sliding action, but also detect duration and pressure related to the touch or sliding operation. The audio element 106 is configured to output and / or input audio signals. For example, the audio element 106 includes a microphone (MIC). When the terminal device is in an operation mode, such as a voice recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 102 or transmitted via the communication element 103. In some embodiments, the audio element 106 further includes a speaker for outputting audio signals. The input / output interface 107 provides an interface between the processing element 100 and a peripheral interface module. The peripheral interface module may be a click wheel, a button, or the like. These buttons may include, but are not limited to, a volume button, a start button, and a lock button. The sensor element 108 includes one or more sensors for providing various aspects of status assessment to the terminal device. For example, the sensor element 108 may detect the on / off state of the terminal device, the relative positioning of the element, the presence or absence of a user's contact with the terminal device. The sensor element 108 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact, including detecting the distance between the user and the terminal device. In some embodiments, the sensor element 108 may further include a camera and the like. The communication element 103 is configured to facilitate wired or wireless communication between the terminal device and other devices. The terminal device can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In one embodiment, the terminal device may include a SIM card slot, and the SIM card slot is used to insert a SIM card, so that the terminal device can log in to the GPRS network and establish communication with the server through the Internet. As can be seen from the above, the communication element 103, the audio element 106, the input / output interface 107, and the sensor element 108 involved in the embodiment of FIG. 12 can be implemented as the input device in the embodiment of FIG. 11. An embodiment of the present application provides a terminal device, including: one or more processors; and one or more machine-readable media having instructions stored thereon, which, when executed by the one or more processors, cause all the The terminal device executes the method for generating a video digest according to one or more of the embodiments of the present application. In one embodiment, the above-mentioned terminal devices may include various smart terminals with computing and processing functions, such as vehicle terminals, mobile terminals (such as mobile phones, tablets, personal digital assistants, etc.), servers, IoT devices, or smart home hardware. Machine equipment. Smart home hardware includes, for example, microwave ovens, ovens, washing machines, dishwashers, air conditioners, routers, smart speakers, TVs, refrigerators, vacuum cleaners, and the like. The above-mentioned smart terminal device may install an application program, provide a human-machine interactive operation interface, and execute the keyword confirmation method of the foregoing embodiments. For example, these smart terminal devices can receive audio data through themselves or an external audio receiving component. After confirming that the second audio data before and after the first audio data is muted, confirm that the first audio data is a valid keyword. For example, for mobile phones, in this way, you can determine whether the voice command issued by the user is a command that instructs an application installed in the phone to perform corresponding operations-such as turning on music, navigation, etc .; for IoT devices or smart home hardware, In this way, it can be determined whether the voice command issued by the user is a command instructing the software or system installed therein to perform corresponding operations-such as connecting other equipment, increasing the temperature of the air conditioner, turning on the high-temperature baking mode of the oven, and the like. It is not particularly limited here. Therefore, it can be known from the foregoing examples that the present invention can be applied to various types of terminal equipment.的 Each embodiment in this specification is described in a progressive manner. Each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. Although the preferred embodiments of the embodiments of the present application have been described, those skilled in the art can make other changes and modifications to these embodiments once they know the basic inventive concepts. Therefore, the scope of the appended application patents is intended to be construed as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of this application. Finally, it should be noted that in this article, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply these entities. There is any such actual relationship or order between OR operations. Moreover, the terms "including", "comprising", or any other variation thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or end device that includes a series of elements includes not only those elements but also those that are not explicitly listed Other elements, or elements inherent to such a process, method, article, or terminal. Without more restrictions, the elements defined by the sentence "including a ..." do not exclude that there are other identical elements in the process, method, article or terminal device including the elements. The keyword confirmation method and device provided in this application have been described in detail above. The specific examples are used in this article to explain the principle and implementation of this application. The description of the above embodiments is only used to help understand this application. Method and its core ideas; meanwhile, for a person of ordinary skill in the art, according to the ideas of the present application, there will be changes in the specific implementation and application scope. In summary, the contents of this specification should not be interpreted as Application restrictions.

90‧‧‧ input device

91‧‧‧ processor

92‧‧‧output device

93‧‧‧Memory

94‧‧‧communication bus

100‧‧‧Processing components

101‧‧‧ processor

102‧‧‧Memory

103‧‧‧Communication components

104‧‧‧Power components

105‧‧‧Multimedia components

106‧‧‧Audio components

107‧‧‧ input / output interface

108‧‧‧Sensor components

200‧‧‧car terminal

300‧‧‧Server

400‧‧‧Speaker

500‧‧‧Internet

600‧‧‧ passengers

700‧‧‧ Microphone

S101, S102, S103, S200, S201, S202, S203, S2021, S2022, S2023, S2024, S2025, S300, S301, S302, S303, S304, S401, S402, S403, S501, S502, S503, S504‧‧‧‧ step

In order to more clearly explain the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are For some embodiments of the invention, for those of ordinary skill in the art, other drawings can be obtained based on these drawings without paying creative labor. FIG. 1 shows a schematic diagram of normal keywords and audio data before and after an embodiment of the present invention. FIG. 2 is a flowchart of a keyword confirmation method according to the first embodiment of the present application. 3 is a flowchart of a keyword confirmation method according to a second embodiment of the present application. FIG. 4 is a sub-step flowchart of the steps in FIG. 3. 5 is a flowchart of a keyword confirmation method according to a third embodiment of the present application. FIG. 6 is a schematic diagram of a vehicle-mounted terminal including a vehicle environment. 7 is a flowchart of a keyword confirmation method for a vehicle-mounted terminal according to a fifth embodiment of the present application. 8 is a flowchart of a keyword confirmation method for a vehicle-mounted terminal according to a sixth embodiment of the present application. 9 is a block diagram of a keyword confirmation apparatus according to a seventh embodiment of the present application. 10 is a block diagram of a keyword confirmation apparatus according to an eighth embodiment of the present application. FIG. 11 schematically shows a block diagram of a terminal device for performing a method according to the present invention. Fig. 12 schematically shows a storage unit for holding or carrying a code for implementing a method according to the present invention.

Claims (20)

A method for confirming a keyword, comprising: acquiring a first audio material, the first audio material being identified as a keyword; determining that there are multiple consecutive mutes in a second audio material that is temporally continuous with the first audio material Fragment; confirm that the first audio material is a valid keyword. The method according to item 1 of the scope of patent application, wherein the step of determining that there are multiple consecutive silence segments in the second audio material that is temporally continuous with the first audio material includes: determining that the pronunciation of the fragment is similar Probability, the pronunciation similarity probability is the maximum probability of similarity between the segment and multiple pronunciation units; determining the silence probability of the segment, the silence probability is the similarity probability between the fragment and the silence unit; when the pronunciation When the relationship between the similarity probability and the mute probability meets a preset condition, the segment is determined to be a mute segment; based on the determined mute segment, it is determined that a plurality of consecutive mute segments exist in the second audio material. The method according to item 1 of the scope of patent application, wherein the step of determining whether at least one segment of the second audio material is a silent segment includes: determining a pronunciation similarity probability of the segment, where the pronunciation similarity probability is The maximum similarity probability between the segment and multiple pronunciation units; determining the silence probability of the segment, the silence probability is the similarity probability between the pronunciation unit and the silence unit corresponding to the maximum similarity probability; when the pronunciation similarity probability and When the relationship of the mute probability meets a preset condition, the segment is determined to be a mute segment; based on the determined mute segment, it is determined that a plurality of consecutive mute segments exist in the second audio material. The method according to item 2 or 3 of the scope of patent application, wherein the step of determining that there are multiple consecutive silence segments in the second audio material includes: determining that there are more than three in the second audio material Continuous mute clips. The method according to item 2 or 3 of the scope of patent application, wherein the preset conditions include: The absolute value of the difference between the pronunciation similarity probability and the silence probability of the segment is less than the first threshold. The method according to item 1 of the scope of patent application, wherein before the step of obtaining audio data, the method further comprises: detecting whether the collected audio data includes keywords. The method according to item 6 of the scope of patent application, wherein the keyword has attribute information, and the step of confirming that the first audio material is a valid keyword includes: When the attribute information of the keyword is the main keyword When the second audio data before the keyword is muted, confirm that the keyword is a valid main keyword. The method according to item 6 of the scope of patent application, wherein the keyword has attribute information, and the step of confirming that the first audio material is a valid keyword includes: When the attribute information of the keyword is a secondary keyword When the second audio material before and after the keyword is mute, it is confirmed that the keyword is a valid sub-keyword. A keyword confirmation method, comprising: ： acquiring a first audio material, said first audio material being identified as a keyword; determining a cumulative mute of a plurality of segments of a second audio material that is temporally continuous with said first audio material Probability; determining the cumulative keyword probability of multiple segments of the first audio material; confirming that the first audio material is valid when the relationship between the cumulative silence probability and the cumulative keyword probability meets a second preset condition word. The method according to item 9 of the scope of patent application, wherein the second preset condition includes: The absolute value of the ratio of the cumulative silence probability to the cumulative keyword probability is greater than a second threshold. The method according to item 9 of the scope of patent application, wherein before the step of obtaining audio data, the method further includes: detecting whether the collected audio data includes keywords. The method according to item 11 of the scope of patent application, wherein the keyword has attribute information, and when the relationship between the cumulative silence probability and the cumulative keyword probability meets a second preset condition, confirming the first audio The step that the data is a valid keyword includes: When the attribute information of the keyword is the main keyword, and the second audio data before the keyword is muted, confirm that the keyword is a valid main keyword. The method according to item 11 of the scope of patent application, wherein the keyword has attribute information, and when the relationship between the cumulative silence probability and the cumulative keyword probability meets a second preset condition, confirming the first audio The step of the data being a valid keyword includes: 确认 When the attribute information of the keyword is a sub-keyword and the second audio data before and after the keyword are both mute, confirm that the keyword is a valid sub-key word. A keyword confirmation method for a vehicle terminal includes: ： acquiring first audio data through a vehicle audio acquisition device, the first audio data being identified as a keyword; determining a second audio time continuous with the first audio data There are multiple consecutive mute segments in the data; Confirm that the first audio data is a valid keyword, and the valid keyword is used to wake up the vehicle terminal to execute the instruction corresponding to the keyword. A keyword confirmation method for a vehicle terminal includes: ： acquiring first audio data through a vehicle audio acquisition device, the first audio data being identified as a keyword; determining a second audio time continuous with the first audio data The cumulative silence probability of multiple segments of the material; determining the cumulative keyword probability of the multiple segments of the first audio material; 确认 when the relationship between the cumulative silence probability and the cumulative keyword probability meets a second preset condition, confirming the The first audio material is a valid keyword, and the valid keyword is used to wake up a vehicle-mounted terminal to execute an instruction corresponding to the keyword. A keyword confirmation device, comprising: an audio data acquisition module for acquiring a first audio data, the first audio data being identified as a keyword; a silent segment determination module for determining a relationship with the first audio There are multiple consecutive mute segments in the second audio material that is continuous in data time; A valid keyword determination module is used to confirm that the first audio material is a valid keyword. A keyword confirmation device, comprising: an audio data acquisition module for acquiring a first audio data, the first audio data being identified as a keyword; a cumulative mute segment determination module for determining a relationship with the first Cumulative silence probability of multiple segments of the second audio material that are continuous in audio material in time; cumulative keyword probability determination module for determining the cumulative keyword probability of multiple fragments of the first audio material; effective keyword determination A module configured to confirm that the first audio data is a valid keyword when the relationship between the cumulative silence probability and the cumulative keyword probability meets a second preset condition. A terminal device includes: (i) one or more processors; and one or more machine-readable media having instructions stored thereon, which when executed by the one or more processors, cause the terminal device to execute a protocol such as Apply one or more of the methods described in claims 1-15. The terminal device according to item 18 of the scope of patent application, wherein the terminal device includes a vehicle terminal, a mobile terminal, a server, an IoT device, or a smart home hardware. One or more machine-readable media having instructions stored thereon, when executed by one or more processors, cause the terminal device to perform the method as described in one or more of items 1-15 of the scope of patent application.